Testing

Writing Tests
Using JUnit
Using TestNG
Excluding Tests and Ignoring Failures
Running Tests
Integration Tests
Using Setup and Teardown
Testing Your Build
Behaviour-Driven Development

Untested code is broken code, so we take testing seriously. Off the bat you get to use either JUnit or TestNG for writing unit tests and integration tests. And you can also add your own framework, or even script tests using Ruby. But first, let’s start with the basics.

Writing Tests

Each project has a TestTask that you can access using the test method. TestTask reflects on the fact that each project has one task responsible for getting the tests to run and acting on the results. But in fact there are several tasks that do all the work, and a test task coordinates all of that.

The first two tasks to execute are test.compile and test.resources. They work similar to compile and resources, but uses a different set of directories. For example, Java tests compile from the src/test/java directory into the target/test/classes directory, while resources are copied from src/test/resources into target/test/resources.

The test.compile task will run the compile task first, then use the same dependencies to compile the test classes. That much you already assumed. It also adds the test framework (e.g. JUnit, TestNG) and JMock to the dependency list. Less work for you.

If you need more dependencies, the best way to add them is by calling test.with. This method adds dependencies to both compile.dependencies (for compiling) and test.dependencies (for running). You can manage these two dependency lists separately, but using test.with is good enough in more cases.

Once compiled, the test task runs all the tests.

Using JUnit

The default test framework for Java projects is JUnit 4.

When you use JUnit, the dependencies includes JUnit, and Buildr picks up all test classes from the project by looking for classes that either subclass junit.framework.TestCase, include methods annotated with org.junit.Test, or test suites annotated with org.org.junit.runner.RunWith.

The JUnit test framework supports the following options:

Option	Value
`:fork`	VM forking, defaults to true.
`:clonevm`	If true clone the VM each time it is forked.
`:properties`	Hash of system properties available to the test case.
`:environment`	Hash of environment variables available to the test case.
`:java_args`	Arguments passed as is to the JVM.

For example, to pass properties to the test case:

test.using :properties=>{ :currency=>'USD' }

There are benefits to running test cases in separate VMs. The default forking mode is :once, and you can change it by setting the :fork option.

:fork=>	Behavior
`:once`	Create one VM to run all test classes in the project, separate VMs for each project.
`:each`	Create one VM for each test case class. Slow but provides the best isolation between test classes.
`false`	Without forking, Buildr runs all test cases in a single VM. This option runs fastest, but at the risk of running out of memory and causing test cases to interfere with each other.

You can see your tests running in the console, and if any tests fail, Buildr will show a list of the failed test classes. In addition, JUnit produces text and XML report files in the project’s reports/junit directory. You can use that to get around too-much-stuff-in-my-console, or when using an automated test system.

In addition, you can get a consolidated XML or HTML report by running the junit:report task. For example:

$ buildr test junit:report test=all
$ firefox report/junit/html/index.html

The junit:report task generates a report from all tests run so far. If you run tests in a couple of projects, it will generate a report only for these two projects. The example above runs tests in all the projects before generating the reports.

Using TestNG

You can also use TestNG in your project by telling your project to use TestNG:

test.using :testng

Like all other options you can set with test.using, it affects the projects and all its sub-projects, so you only need to do this once at the top-most project to use TestNG throughout. You can also mix TestNG and JUnit by setting different projects to use different frameworks, but you can’t mix both frameworks in the same project. (And yes, test.using :junit will switch a project back to using JUnit)

TestNG works much like JUnit, it gets included in the dependency list, Buildr picks test classes that contain methods annotated with org.testng.annotations.Test, and generates test reports in the reports/testng directory. At the moment we don’t have consolidated HTML reports for TestNG.

The TestNG test framework supports the following options:

Option	Value
`:properties`	Hash of system properties available to the test case.
`:java_args`	Arguments passed as is to the JVM.

Excluding Tests and Ignoring Failures

If you have a lot of tests that are failing or just hanging there collecting dusts, you can tell Buildr to ignore them. You can either tell Buildr to only run specific tests, for example:

test.include 'com.acme.tests.passing.*'

Or tell it to exclude specific tests, for example:

test.exclude '*FailingTest', '*FailingWorseTest'

Note that we’re always using the package qualified class name, and you can use star (*) to substitute for any set of characters.

When tests fail, Buildr fails the test task. This is usually a good thing, but you can also tell Buildr to ignore failures by resetting the :fail_on_failure option:

test.using :fail_on_failure=>false

Besides giving you a free pass to ignore failures, you can use it for other causes, for example, to be somewhat forgiving:

test do
  fail 'More than 3 tests failed!' if test.failed_tests.size > 3
end

The failed_tests collection holds the names of all classes with failed tests. And there’s classes, which holds the names of all test classes. Ruby arithmetic allows you to get the name of all passed test classes with a simple test.classes – test.failed_tests. We’ll let you imagine creative use for these two.

Running Tests

It’s a good idea to run tests every time you change the source code, so we wired the build task to run the test task at the end of the build. And conveniently enough, the build task is the default task, so another way to build changes in your code and run your tests:

$ buildr

That only works with the local build task and any local task that depends on it, like package, install and upload. Each project also has its own build task that does not invoke the test task, so buildr build will run the tests cases, but buildr foo:build will not.

While it’s a good idea to always run your tests, it’s not always possible. There are two ways you can get build to not run the test task. You can set the environment variable test to no (but skip and off will also work). You can do that when running Buildr:

$ buildr test=no

Or set it once in your environment:

$ export TEST=no
$ buildr

If you’re feeling really adventurous, you can also disable tests from your Buildfile or buildr.rb file, by setting options.test = false. We didn’t say it’s a good idea, we’re just giving you the option.

The test task is just smart enough to run all the tests it finds, but will accept include/exclude patterns. Often enough you’re only working on one broken test and you only want to run that one test. Better than changing your Buildfile, you can run the test task with a pattern. For example:

$ buildr test:KillerAppTest

Buildr will then run only tests that match the pattern KillerAppTest. It uses pattern matching, so test:Foo will run com.acme.FooTest and com.acme.FooBarTest. With Java, you can use this to pick a class name, or a package name to run all tests in that package, or any such combination. In fact, you can specify several patterns separated with commas. For example:

$ buildr test:FooTest,BarTest

As you probably noticed, Buildr will stop your build at the first test that fails. We think it’s a good idea, except when it’s not. If you’re using a continuous build system, you’ll want a report of all the failed tests without stopping at the first failure. To make that happen, set the environment variable test to “all”, or the Buildr options.test option to :all. For example:

$ buildr package test=all

We’re using package and not build above. When using a continuous build system, you want to make sure that packages are created, contain the right files, and also run the integration tests.

Integration Tests

So far we talked about unit tests. Unit tests are run in isolation on the specific project they test, in an isolated environment, generally with minimal setup and teardown. You get a sense of that when we told you tests run after the build task, and include JMock in the dependency list.

In contrast, integration tests are run with a number of components, in an environment that resembles production, often with more complicates setup and teardown procedures. In this section we’ll talk about the differences between running unit and integration tests.

You write integration tests much the same way as you write unit tests, using test.compile and test.resources. However, you need to tell Buildr that your tests will execute during integration test. To do so, add the following line in your project definition:

test.using :integration

Typically you’ll use unit tests in projects that create internal modules, such as JARs, and integration tests in projects that create components, such as WARs and EARs. You only need to use the :integration option with the later.

To run integration tests on the current project:

$ buildr integration

You can also run specific tests cases, for example:

$ buildr integration:ClientTest

If you run the package task (or any task that depends on it, like install and upload), Buildr will first run the build task and all its unit tests, and then create the packages and run the integration tests. That gives you full coverage for your tests and ready to release packages. As with unit tests, you can set the environment variable test to “no” to skip integration tests, or “all” to ignore failures.

Using Setup and Teardown

Some tests need you to setup an environment before they run, and tear it down afterwards. The test frameworks (JUnit, TestNG) allow you to do that for each test. Buildr provides two additional mechanisms for dealing with more complicated setup and teardown procedures.

Integration tests run a setup task before the tests, and a teardown task afterwards. You can use this task to setup a Web server for testing your Web components, or a database server for testing persistence. You can access either task by calling integration.setup and integration.teardown. For example:

integration.setup { server.start ; server.deploy }
integration.teardown { server.stop }

Depending on your build, you may want to enhance the setup/teardown tasks from within a project, for example, to populate the database with data used by that project’s test, or from outside the project definition, for example, to start and stop the Web server.

Likewise, each project has its own setup and teardown tasks that are run before and after tests for that specific project. You can access these tasks using test.setup and test.teardown.

Testing Your Build

So you got the build running and all the tests pass, binaries are shipping when you find out some glaring omissions. The license file is empty, the localized messages for Japanese are missing, the CSS files are not where you expect them to be. The fact is, some errors slip by unit and integration tests. So how do we make sure the same mistake doesn’t happen again?

Each project has a check task that runs just after packaging. You can use this task to verify that your build created the files you wanted it to create. And to make it extremely convenient, we introduced the notion of expectations.

You use the check method to express and expectation. Buildr will then run all these expectations against your project, and fail at the first expectation that doesn’t match. An expectation says three things. Let’s look at a few examples:

check package(:war), 'should exist' do
  it.should exist
end
check package(:war), 'should contain a manifest' do
  it.should contain('META-INF/MANIFEST.MF')
end
check package(:war).path('WEB-INF'), 'should contain files' do
  it.should_not be_empty
end
check package(:war).path('WEB-INF/classes'), 'should contain classes' do 
  it.should contain('**/*.class')
end
check package(:war).entry('META-INF/MANIFEST'), 'should have license' do
  it.should contain(/Copyright (C) 2007/)
end
check file('target/classes'), 'should contain class files' do
  it.should contain('**/*.class')
end
check file('target/classes/killerapp/Code.class'), 'should exist' do
  it.should exist
end

The first argument is the subject, or the project if you skip the first argument. The second argument is the description, optional, but we recommend using it. The method it returns the subject.

You can also write the first expectation like this:

check do
  package(:jar).should exist
end

We recommend using the subject and description, they make your build easier to read and maintain, and produce better error messages.

There are two methods you can call on just about any object, called should and should_not. Each method takes an argument, a matcher, and executes that matcher. If the matcher returns false, should fails. You can figure out what should_not does in the same case.

Buildr provides the following matchers:

Method	Checks that …
`exist`	Given a file task checks that the file (or directory) exists.
`empty`	Given a file task checks that the file (or directory) is empty.
`contain`	Given a file task referencing a file, checks its contents, using string or regular expression. For a file task referencing a directory, checks that it contains the specified files; global patterns using `` and `*` are allowed.

All these matchers operate against a file task. If you run them against a ZipTask (including JAR, WAR, etc) they can also check the contents of the ZIP file. And as you can see in the examples above, you can also run them against a path in a ZIP file, checking its contents as if it was a directory, or against an entry in a ZIP file, checking the content of that file.

The package method returns a package task based on packaging type, identifier, group, version and classifier. The last four are inferred, but if you create a package with different specifications (for example, you specify a classifier) your checks must call package with the same qualifying arguments to return the very same package task.

Buildr expectations are based on RSpec. RSpec is the behavior-driven development framework we use to test Buildr itself. Check the RSpec documentation if want to see all the supported matchers, or want to write your own.

Behaviour-Driven Development

Buildr supports several Behaviour-Driven Development(BDD) frameworks for testing your projects. Buildr follows each framework naming conventions, searching for files under the src/spec/{lang} directory.

JBehave

JBehave is a pure Java BDD framework, stories and behaviour specifications are written in the Java language.

To use JBehave in your project you can select it with test.using :jbehave.

This framework will search for the following patterns under your project:

src/spec/java/**/*Behaviour.java

Supports the following options:

Option	Value
`:properties`	Hash of system properties available to the test case.
`:java_args`	Arguments passed as is to the JVM.

RSpec

RSpec is the de-facto BDD framework for ruby. It’s the framework used to test Buildr itself.

Specifications are written in Ruby language, but are run by using JRuby. That means you have access to all your Java classes and any Java or Ruby tool out there.

To use this framework in your project you can select it with @test.using :rspec@.

This framework will search for the following patterns under your project:

src/spec/ruby/**/*_spec.rb

Supports the following options:

Option	Value
`:properties`	Hash of system properties available to the test case.
`:java_args`	Arguments passed as is to the JVM.

EasyB

EasyB is a BDD framework using Groovy.

Specifications are written in the Groovy language, of course you get seamless Java integration as with all things groovy.

To use this framework in your project you can select it with @test.using :easyb@.

This framework will search for the following patterns under your project:

src/spec/groovy/**/*Behavior.groovy
src/spec/groovy/**/*Story.groovy

Supports the following options:

Option	Value
`:properties`	Hash of system properties available to the test case.
`:java_args`	Arguments passed as is to the JVM.
`:format`	Report format, either `:txt` or `:xml`

Next, let’s talk about customizing your environment and using profiles